1. The new UN interagency maternal
mortality estimates
Agbessi Amouzou and Holly Newby
Data & Analytics Section, DPS, UNICEF
1 May 2014
2. Will be released on Tuesday, 6
May 2014
Levels and trends of maternal
mortality between 1990 and
2013 for 183 countries
Includes MMR, lifetime risk of
maternal death and numbers of
maternal deaths
Will replace current UN
interagency estimates pertaining
to 2010
3. Outline of the Presentation
Why UN Inter-Agency estimate of maternal
mortality
Summary of issues in maternal mortality
measurement
How the UN Inter-Agency estimates are
produced
Highlights of new estimates
5. Why Inter-Agency MM Estimates?
MMR is key indicator for MDG 5
Global monitoring and reporting requires a harmonized
measure of MMR that is comparable across countries
Need to obtain a measure that has same reference year
across all countries
Maternal mortality is challenging to measure
Similar initiative is done for under-five mortality (see
www.childmortality.org)
6. Maternal Mortality Estimation
Interagency Group (MMEIG)
The UN interagency estimates are produced by
the Maternal Mortality Estimation Interagency
Group (MMEIG):
• WHO (Lead)
• UNICEF
• UNFPA
• The World Bank
• Lead technical consultant (Leontine Alkema, National
University of Singapor)
• Technical Advisory Group
8. Definitions
Maternal death
The death of a woman while pregnant
or within 42 days of termination of
pregnancy, irrespective of the duration
and site of the pregnancy, from any
cause related to or aggravated by the
pregnancy or its management but not
from accidental or incidental causes.
Death must be attributed directly or
indirectly to pregnancy or childbirth
Requires medical certification or
verbal autopsy
Cannot be obtained through surveys or
censuses
No deaths beyond 42 days due to
pregnancy complications accounted for
Definition Implications
Pregnancy-related death
The death of a woman while pregnant
or within 42 days of termination of
pregnancy, irrespective of the cause of
death.
Cause of death certification not
needed
Can be obtained through surveys
or censuses
UN Interagency maternal mortality estimates conform to the definition of maternal death
9. Sources of maternal mortality data and
their limitations
Maternal mortality data can come from a
variety of sources:
– Vital registration • Considered gold standard
• Good in only about a third of
countries
• Extensive under-reporting
and misclassification
• Even in countries with
complete vital registration,
maternal deaths may be
underreported by a factor of
1.5 – 3.0
10. Sources of maternal mortality data and
their limitations
Maternal mortality data can come from a
variety of sources:
– Vital registration
– Household surveys (sisterhood method)
• Pregnancy-related deaths
• MMR very imprecise, large
confidence intervals
• Doe not produce recent
estimate: MMR refers to 7 to
9 years in the past
11. Sources of maternal mortality data and
their limitations
Maternal mortality data can come from a
variety of sources:
– Vital registration
– Household surveys (sisterhood method)
– Censuses
• Pregnancy-related deaths
• Conducted every 10 years
• Need adjustment for completeness
of births and deaths
12. Sources of maternal mortality data and
their limitations
Maternal mortality data can come from a
variety of sources:
– Vital registration
– Household surveys (sisterhood method, etc.)
– Censuses
– Reproductive-age mortality studies (RAMOS)
• Complicate, time-consuming and
expensive
• Under-report of maternal deaths
• Under report of number of live
births
13. Sources of maternal mortality data and
their limitations
Maternal mortality data can come from a
variety of sources:
– Vital registration
– Household surveys (sisterhood method, etc.)
– Censuses
– Reproductive-age mortality studies (RAMOS)
– Verbal autopsy • Misclassification of cause of
death
• Under report of maternal
deaths
• Recall issues
14. Sources of maternal mortality data and
their limitations
Maternal mortality data can come from a
variety of sources:
– Vital registration
– Household surveys (sisterhood method, etc.)
– Censuses
– Reproductive-age mortality studies (RAMOS)
– Verbal autopsy
Bottom line:
Each source has advantages and limitations.
Measurement is challenging regardless of
source.
There is need to adjust and harmonize
available data for cross country
comparability and global reporting
15. Issues to keep in mind
Survey estimates of MMR are averages over periods
of 7 or 9 years in the past, so not comparable to UN
Interagency estimates
MMR generally have large uncertainty ranges
Maternal death is a rare event; MMR is expressed in
per 100,000 live births and therefore creates a false
sense of precision
– 300/100,000 = 0.30/100
– 330/100,000 = 0.33/100
MMR of 300 may not
be different from
MMR of 330
16. Trend Estimation from Sibling Histories
with 95% Confidence Intervals (Namibia)
Estimates are averages over long periods (here 7 or 9 years) and 95%
confidence intervals are large
0
100
200
300
400
500
600
1985 1990 1995 2000 2005
Year
1992 DHS 2000 DHS
2007 DHS
Source: Ken Hill – UN maternal mort workshop, Nairobi December 2010
The 2007 MMR
refers to period
1998 -2007
17. Trend Estimation from Sibling Histories
with 95% Confidence Intervals (Namibia)
Estimates are averages over long periods (here 7 or 9 years) and 95%
confidence intervals are large
0
100
200
300
400
500
600
1985 1990 1995 2000 2005
Year
1992 DHS 2000 DHS
2007 DHS
Source: Ken Hill – UN maternal mort workshop, Nairobi December 2010
The 2000 MMR has
95%CI ranging from
90 to 450
18. Trend Estimation from Sibling Histories
with 95% Confidence Intervals (Namibia)
Estimates are averages over long periods (here 7 or 9 years) and 95%
confidence intervals are large
0
100
200
300
400
500
600
1985 1990 1995 2000 2005
Year
1992 DHS 2000 DHS
2007 DHS
Source: Ken Hill – UN maternal mort workshop, Nairobi December 2010
Note that this is at the
national level! It’s not possible
to disaggregate by region or
other characteristics like
household wealth!
20. Source of data for the 2013 MMR estimates
Group Source of maternal
mortality data
Number of
countries/
territories
% of
countries/
territories in
each category
% of births
in 183
countries/te
rritories
covered
A Civil registration
characterized as complete,
with good attribution of
cause of death
67 37 17
B Incomplete civil registration
and/or other types of data
96 52 81
C No national data on
maternal mortality
20 11 2
Total 183 100 100
21. General methodology of estimation
Little change from methodology used
for 2010 estimates
1. Compile and review all available nationally
representative maternal mortality data
2. Adjust available maternal mortality data for
misclassification and underreporting
22. 3. Use one of two approaches depending on country
– Countries with adequate civil registration data
• Calculate MMR directly with adjusted
– All other countries:
• Use multilevel linear regression model
• Covariates: GDP, general fertility rate and skilled attendant at
birth
• Separate model component for AIDS deaths that are indirect
maternal deaths
4. Compute uncertainty ranges through simulations
General methodology of estimation
23. Methodological changes from the
2010 estimates?
Increased data availability
– 5% increase in available data
Update in the estimate of female deaths in the
reproductive age by WHO
Update of series of live births and general
fertility rates from World Population Prospects
Update in AIDS adjustment parameters
24. Methodological changes from the
2010 estimates?
Data availability
– 5% increase in available data
Update in the estimate of female deaths in the
reproductive age by WHO
Update of series of live births and general
fertility rates from World Population Prospects
Update in AIDS adjustment parameters
Little change from methodology used
for 2010 estimates
25. Review process
Reviewed by the Technical Advisory Group
with experts from academic institutions:
Harvard University, Johns Hopkins University,
University of Aberdeen, and others
Country consultation led by WHO allowed
countries to provide feedback and provide
new data
26. STOP!
The 2013 UN interagency estimates REPLACE
the previous estimates and should not be
compared or interpreted together with them
The 2013 estimates are NOT comparable to
estimates from other sources
27. Maternal mortality estimates generated by
countries
At the global level, we use the interagency estimates for MDG
reporting and official monitoring
UNICEF presents both nationally reported estimates and UN
interagency estimates in State of the World’s Children
TABLE 8
28. Embargoed until May 6, 2014 ----------------------
990
550
200
140
170
65
940
430
12
380
680
280
140
93
100
36
610
300
11
270
510
190
110
85
74
27
440
230
15
210
0
200
400
600
800
1000
1200
1990 2005 2013
Trends in Maternal Mortality Ratio
(Embargoed until May 6, 2014)
By UNICEF regions
Source: Trends in Maternal Mortality: 1990-2013 (WHO, UNICEF, UNFPA, World Bank)
29. Resources
Complete methodological details and all data available on:
www.who.int/reproductivehealth/publications/monitoring/xxxxxxxxx/en/ind
ex.html and MME Info: www.maternalmortalitydata.org
More information on new estimates available (from
May 6) at:
Data.unicef.org
We are in process of updating the MMEIG website MM Info
(maternalmortalitydata.org)
30. To be released on
6 May 2014!
Contacts
Agbessi Amouzou
aamouzou@unicef.org
Holly Newby
hnewby@unicef.org
Editor's Notes
This inter-agency group began working together in the mid-1990s with the goal of providing a more accurate assessment of the global maternal mortality burden, as well as comparable estimates across countries. The MMEIG has produced peer reviewed sets of estimates that have been critical for MDG5 monitoring and reporting.
Vital registration: Considered gold standard, however….
Relatively few countries have complete vital registration and good attribution of cause of death
Extensive under-reporting and misclassification
Even in countries with complete vital registration, maternal deaths may be underreported by a factor of 1.5 – 3.0
Vital registration: Considered gold standard, however….
Relatively few countries have complete vital registration and good attribution of cause of death
Extensive under-reporting and misclassification
Even in countries with complete vital registration, maternal deaths may be underreported by a factor of 1.5 – 3.0
Household surveys (sisterhood method): Only source of information in many developing countries, however….
Estimates refer to a period 0-6 or 0-9 years before the survey
Wide confidence intervals
Household surveys (sisterhood method): Only source of information in many developing countries, however….
Estimates refer to a period 0-6 or 0-9 years before the survey
Wide confidence intervals
Vital registration: Considered gold standard, however….
Relatively few countries have complete vital registration and good attribution of cause of death
Extensive under-reporting and misclassification
Even in countries with complete vital registration, maternal deaths may be underreported by a factor of 1.5 – 3.0
Household surveys (sisterhood method): Only source of information in many developing countries, however….
Estimates refer to a period 0-6 or 0-9 years before the survey
Wide confidence intervals
Vital registration: Considered gold standard, however….
Relatively few countries have complete vital registration and good attribution of cause of death
Extensive under-reporting and misclassification
Even in countries with complete vital registration, maternal deaths may be underreported by a factor of 1.5 – 3.0
Household surveys (sisterhood method): Only source of information in many developing countries, however….
Estimates refer to a period 0-6 or 0-9 years before the survey
Wide confidence intervals
Vital registration: Considered gold standard, however….
Relatively few countries have complete vital registration and good attribution of cause of death
Extensive under-reporting and misclassification
Even in countries with complete vital registration, maternal deaths may be underreported by a factor of 1.5 – 3.0
Household surveys (sisterhood method): Only source of information in many developing countries, however….
Estimates refer to a period 0-6 or 0-9 years before the survey
Wide confidence intervals